Scaling Laws for the Few-Shot Adaptation of Pre-trained Image Classifiers